scholarly journals Forecasting of Meteorologically Driven “Extremes” in Wind and Solar Power: Can We Tackle and Improve Selected Cases of Non-forecasted Extreme Events using Deep Learning?

2021 ◽  
Author(s):  
Petrina Papazek ◽  
Irene Schicker ◽  
Rosmarie de Wit
Keyword(s):  
Renewable Energy ◽  
Deep Learning ◽  
Extreme Events ◽  
Energy Production ◽  
Solar Power ◽  
Prediction Models ◽  
Weather Prediction ◽  
Power Production ◽  
Saharan Dust ◽  
Learning Framework

<p>With the rapid transition towards an increased usage of renewable energy accurate predictions of the expected power production are needed to ensure grid stability, energy trading, and (re)scheduling of maintenance and energy transfer. In the last decades, both numerical weather prediction models as well as post-processing methods have significantly improved their prediction skills when applied to renewable energy. There are, however, events in renewable energy production which can be considered as extreme events but are not necessarily extremes in terms of meteorological conditions. The MEDEA project, funded by the Austrian Climate Research Program, aims at improving the definition and detection of extreme events relevant for renewable energies and to use these findings to improve both weather and climate predictions of such extreme events. In this MEDEA case study, we investigate selected (extremes) cases in renewable energy which were not properly reproduced by the models. We will have a deeper dive into two Saharan dust events in 2021 where none of the models was able to properly reproduce the amount of aerosols in Central Europe and solar power production was off by more than 5 GW in contrast to the predictions.  Here, several NWP models have failed to properly recognize its impact and, thus, impaired results based on raw model output. To tackle such events and improve the predictability, a deep learning framework consisting of an auto-encoder LSTM (long short-term memory; type of an artificial neural network) and random forest will be used and adopted for day-ahead predictions of these events. Relevant features for the learning algorithms are extracted from different NWP models, satellite data, and observations. Similarly, for wind energy production we demonstrate the methods in two selected case studies of extreme events. Results obtained by the deep learning framework yield, in general, high forecast-skills where we elaborate on interesting cases studies from a meteorological point of view. Different combinations of inputs and processing-steps are part of the analysis. We compare results to traditional forecast methods in order to validate the performance of our methods.</p>

scholarly journals Enabling deeper learning on big data for materials informatics applications

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Dipendra Jha ◽  
Vishu Gupta ◽  
Logan Ward ◽  
Zijiang Yang ◽  
Christopher Wolverton ◽  
...  
Keyword(s):  
Neural Networks ◽  
Big Data ◽  
Deep Learning ◽  
Deep Neural Networks ◽  
Materials Science ◽  
Prediction Models ◽  
Model Performance ◽  
Materials Informatics ◽  
Learning Framework ◽  
Significant Attention

AbstractThe application of machine learning (ML) techniques in materials science has attracted significant attention in recent years, due to their impressive ability to efficiently extract data-driven linkages from various input materials representations to their output properties. While the application of traditional ML techniques has become quite ubiquitous, there have been limited applications of more advanced deep learning (DL) techniques, primarily because big materials datasets are relatively rare. Given the demonstrated potential and advantages of DL and the increasing availability of big materials datasets, it is attractive to go for deeper neural networks in a bid to boost model performance, but in reality, it leads to performance degradation due to the vanishing gradient problem. In this paper, we address the question of how to enable deeper learning for cases where big materials data is available. Here, we present a general deep learning framework based on Individual Residual learning (IRNet) composed of very deep neural networks that can work with any vector-based materials representation as input to build accurate property prediction models. We find that the proposed IRNet models can not only successfully alleviate the vanishing gradient problem and enable deeper learning, but also lead to significantly (up to 47%) better model accuracy as compared to plain deep neural networks and traditional ML techniques for a given input materials representation in the presence of big data.

scholarly journals Understanding the relationship between clouds and surface downward radiation forecast errors with Unsupervised Deep Learning

2021 ◽  
Author(s):  
Matthias Zech ◽  
Lueder von Bremen
Keyword(s):  
Deep Learning ◽  
Unsupervised Learning ◽  
Prediction Models ◽  
Weather Prediction ◽  
Reanalysis Data ◽  
Cloud Formation ◽  
Trade Wind ◽  
Forecast Errors ◽  
Model Learning ◽  
Unsupervised Deep Learning

<p>Cloudiness is a difficult parameter to forecast and has improved relatively little over the last decade in numerical weather prediction models as the EMCWF IFS. However, surface downward solar radiation forecast (ssrd) errors are becoming more important with higher penetration of photovoltaics in Europe as forecasts errors induce power imbalances that might lead to high balancing costs. This study continues recent approaches to better understand clouds using satellite images with Deep Learning. Unlike other studies which focus on shallow trade wind cumulus clouds over the ocean, this study investigates the European land area. To better understand the clouds, we use the daily MODIS optical cloud thickness product which shows both water and ice phase of the cloud. This allows to consider both cloud structure and cloud formation during learning. It is also much easier to distinguish between snow and cloud in contrast to using visible bands. Methodologically, it uses the Unsupervised Learning approach <em>tile2vec</em> to derive a lower dimensional representation of the clouds. Three cloud regions with two similar neighboring tiles and one tile from a different time and location are sampled to learn lower-rank embeddings. In contrast to the initial <em>tile2vec</em> implementation, this study does not sample arbitrarily distant tiles but uses the fractal dimension of the clouds in a pseudo-random sampling fashion to improve model learning.</p><p>The usefulness of the cloud segments is shown by applying them in a case study to investigate statistical properties of ssrd forecast errors over Europe which are derived from hourly ECMWF IFS forecasts and ERA5 reanalysis data. This study shows how Unsupervised Learning has high potential despite its relatively low usage compared to Supervised Learning in academia. It further shows, how the generated land cloud product can be used to better characterize ssrd forecast errors over Europe.</p>

scholarly journals A Variable-Correlation Model to Characterize Asymmetric Dependence for Postprocessing Short-Term Precipitation Forecasts

2019 ◽  
Vol 148 (1) ◽  
pp. 241-257 ◽  
Author(s):  
Wentao Li ◽  
Quan J. Wang ◽  
Qingyun Duan
Keyword(s):  
Correlation Coefficient ◽  
Extreme Events ◽  
Prediction Models ◽  
Probability Model ◽  
Weather Prediction ◽  
Joint Probability ◽  
Correlation Model ◽  
Asymmetric Dependence ◽  
Huai River ◽  
Joint Probability Model

Abstract Statistical postprocessing methods can be used to correct bias and dispersion error in raw ensemble forecasts from numerical weather prediction models. Existing postprocessing models generally perform well when they are assessed on all events, but their performance for extreme events still needs to be investigated. Commonly used joint probability postprocessing models are based on the correlation between forecasts and observations. Because the correlation may be lower for extreme events as a result of larger forecast uncertainty, the dependence between forecasts and observations can be asymmetric with respect to the magnitude of the precipitation. However, the constant correlation coefficient in the traditional joint probability model lacks the flexibility to model asymmetric dependence. In this study, we formulated a new postprocessing model with a decreasing correlation coefficient to characterize asymmetric dependence. We carried out experiments using Global Ensemble Forecast System reforecasts for daily precipitation in the Huai River basin in China. The results show that, although it performs well in terms of continuous ranked probability score or reliability for all events, the traditional joint probability model suffers from overestimation for extreme events defined by the largest 2.5% or 5% of raw forecasts. On the contrary, the proposed variable-correlation model is able to alleviate the overestimation and achieves better reliability for extreme events than the traditional model. The proposed variable-correlation model can be seen as a flexible extension of the traditional joint probability model to improve the performance for extreme events.

scholarly journals Short-Term Forecasting of Photovoltaic Solar Power Production Using Variational Auto-Encoder Driven Deep Learning Approach

2020 ◽  
Vol 10 (23) ◽  
pp. 8400 ◽  
Author(s):  
Abdelkader Dairi ◽  
Fouzi Harrou ◽  
Ying Sun ◽  
Sofiane Khadraoui
Keyword(s):  
Machine Learning ◽  
Deep Learning ◽  
Solar Power ◽  
Power Production ◽  
Superior Performance ◽  
Support Vector ◽  
Learning Models ◽  
Short Term ◽  
Learning Methods ◽  
Short Term Forecasting

The accurate modeling and forecasting of the power output of photovoltaic (PV) systems are critical to efficiently managing their integration in smart grids, delivery, and storage. This paper intends to provide efficient short-term forecasting of solar power production using Variational AutoEncoder (VAE) model. Adopting the VAE-driven deep learning model is expected to improve forecasting accuracy because of its suitable performance in time-series modeling and flexible nonlinear approximation. Both single- and multi-step-ahead forecasts are investigated in this work. Data from two grid-connected plants (a 243 kW parking lot canopy array in the US and a 9 MW PV system in Algeria) are employed to show the investigated deep learning models’ performance. Specifically, the forecasting outputs of the proposed VAE-based forecasting method have been compared with seven deep learning methods, namely recurrent neural network, Long short-term memory (LSTM), Bidirectional LSTM, Convolutional LSTM network, Gated recurrent units, stacked autoencoder, and restricted Boltzmann machine, and two commonly used machine learning methods, namely logistic regression and support vector regression. The results of this investigation demonstrate the satisfying performance of deep learning techniques to forecast solar power and point out that the VAE consistently performed better than the other methods. Also, results confirmed the superior performance of deep learning models compared to the two considered baseline machine learning models.

Comparison of Single-Valued Forecasts in a User-Oriented Framework

2020 ◽  
Vol 35 (3) ◽  
pp. 1067-1080
Author(s):  
Michael Foley ◽  
Nicholas Loveday
Keyword(s):  
Numerical Weather Prediction ◽  
Extreme Events ◽  
Prediction Models ◽  
Weather Prediction ◽  
Economic Value ◽  
Extreme Conditions ◽  
Single Model ◽  
Consensus Forecasts ◽  
Theoretical Results ◽  
Numerical Weather Prediction Models

Abstract We compare single-valued forecasts from a consensus of numerical weather prediction models to forecasts from a single model across a range of user decision thresholds and sensitivities, using the relative economic value framework, and present this comparison in a new graphical format. With the help of a simple linear error model, we obtain theoretical results and perform synthetic calculations to gain insights into how the results relate to the characteristics of the different forecast systems. We find that multimodel consensus forecasts are more beneficial for users interested in decisions near the climatological mean, due to their reduced spread of errors compared to the constituent models. Single model forecasts may present greater benefit for users sensitive to extreme events if the forecasts have smaller conditional biases than the consensus forecasts and hence better resolution of such events. The results support use of consensus averaging approaches for single-valued forecast services in typical conditions. However, it is hard to cater for all user sensitivities in more extreme conditions. This underscores the importance of providing probability-based services for unusual conditions.

Forecasting the Wind to Reach Significant Penetration Levels of Wind Energy

2011 ◽  
Vol 92 (9) ◽  
pp. 1159-1171 ◽  
Author(s):  
Melinda Marquis ◽  
Jim Wilczak ◽  
Mark Ahlstrom ◽  
Justin Sharp ◽  
Andrew Stern ◽  
...  
Keyword(s):  
Renewable Energy ◽  
Wind Energy ◽  
Prediction Models ◽  
Weather Prediction ◽  
Atmospheric Science ◽  
Time Frame ◽  
Department Of Energy ◽  
Power Prediction ◽  
Financial Benefit ◽  
The Government

Advances in atmospheric science are critical to increased deployment of variable renewable energy (VRE) sources. For VRE sources, such as wind and solar, to reach high penetration levels in the nation's electric grid, electric system operators and VRE operators need better atmospheric observations, models, and forecasts. Improved meteorological observations through a deep layer of the atmosphere are needed for assimilation into numerical weather prediction (NWP) models. The need for improved operational NWP forecasts that can be used as inputs to power prediction models in the 0–36-h time frame is particularly urgent and more accurate predictions of rapid changes in VRE generation (ramp events) in the very short range (0–6 h) are crucial. We describe several recent studies that investigate the feasibility of generating 20% or more of the nation's electricity from weather-dependent VRE. Next, we describe key advances in atmospheric science needed for effective development of wind energy and approaches to achieving these improvements. The financial benefit to the nation of improved wind forecasts is potentially in the billions of dollars per year. Obtaining the necessary meteorological and climatological observations and predictions is a major undertaking, requiring collaboration from the government, private, and academic sectors. We describe a field project that will begin in 2011 to improve short-term wind forecasts, which demonstrates such a collaboration, and which falls under a recent memorandum of understanding between the Office of Energy Efficiency and Renewable Energy at the Department of Energy and the Department of Commerce/National Oceanic and Atmospheric Administration.

scholarly journals kGCN: A Graph-Based Deep Learning Framework for Chemical Structures

2020 ◽  
Author(s):  
Ryosuke Kojima ◽  
Shoichi Ishida ◽  
Masateru Ohta ◽  
Hiroaki Iwata ◽  
Teruki Honma ◽  
...  
Keyword(s):  
Deep Learning ◽  
Prediction Model ◽  
Prediction Models ◽  
Bayesian Optimization ◽  
Factors Affecting ◽  
Learning Framework ◽  
Chemical Structures ◽  
Programming Skills ◽  
Model Tuning ◽  
Interpretation Of Results

<div>Deep learning is developing as an important technology to perform various tasks in cheminformatics. In particular, graph convolutional neural networks (GCNs) have been reported to perform well in many types of prediction tasks related to molecules. Although GCN exhibits considerable potential in various applications, appropriate utilization of this resource for obtaining reasonable and reliable prediction results requires thorough understanding of GCN and programming. To leverage the power of GCN to benefit various users from chemists to cheminformaticians, an open-source GCN tool, kGCN, is introduced. To support the users with various levels of programming skills, kGCN includes three interfaces: a graphical user interface (GUI) employing KNIME for users with limited programming skills such as chemists, as well as command-line and Python library interfaces for users with advanced programming skills such as cheminformaticians. To support the three steps required for building a prediction model, i.e., pre-processing, model tuning, and interpretation of results, kGCN includes functions of typical pre-processing, Bayesian optimization for automatic model tuning, and visualization of the atomic contribution to prediction for interpretation of results. kGCN supports three types of approaches, single-task, multi-task, and multimodal predictions. The prediction of compound-protein interaction for four matrixmetalloproteases, MMP-3, -9, -12 and -13, in the inhibition assays is performed as a representative case study using kGCN. Additionally, kGCN provides the visualization of atomic contributions to the prediction. Such visualization is useful for the validation of the prediction models and the design of molecules based on the prediction model, realizing “explainable AI” for understanding the factors affecting AI prediction. kGCN is available at https://github.com/clinfo/kGCN.</div>

scholarly journals RPITER: A Hierarchical Deep Learning Framework for ncRNA–Protein Interaction Prediction

2019 ◽  
Vol 20 (5) ◽  
pp. 1070 ◽  
Author(s):  
Cheng Peng ◽  
Siyu Han ◽  
Hui Zhang ◽  
Ying Li
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Protein Interaction ◽  
Rna Binding ◽  
Prediction Models ◽  
Rna Binding Proteins ◽  
Biological Research ◽  
Sequence Information ◽  
Learning Framework ◽  
Coding Method

Non-coding RNAs (ncRNAs) play crucial roles in multiple fundamental biological processes, such as post-transcriptional gene regulation, and are implicated in many complex human diseases. Mostly ncRNAs function by interacting with corresponding RNA-binding proteins. The research on ncRNA–protein interaction is the key to understanding the function of ncRNA. However, the biological experiment techniques for identifying RNA–protein interactions (RPIs) are currently still expensive and time-consuming. Due to the complex molecular mechanism of ncRNA–protein interaction and the lack of conservation for ncRNA, especially for long ncRNA (lncRNA), the prediction of ncRNA–protein interaction is still a challenge. Deep learning-based models have become the state-of-the-art in a range of biological sequence analysis problems due to their strong power of feature learning. In this study, we proposed a hierarchical deep learning framework RPITER to predict RNA–protein interaction. For sequence coding, we improved the conjoint triad feature (CTF) coding method by complementing more primary sequence information and adding sequence structure information. For model design, RPITER employed two basic neural network architectures of convolution neural network (CNN) and stacked auto-encoder (SAE). Comprehensive experiments were performed on five benchmark datasets from PDB and NPInter databases to analyze and compare the performances of different sequence coding methods and prediction models. We found that CNN and SAE deep learning architectures have powerful fitting abilities for the k-mer features of RNA and protein sequence. The improved CTF coding method showed performance gain compared with the original CTF method. Moreover, our designed RPITER performed well in predicting RNA–protein interaction (RPI) and could outperform most of the previous methods. On five widely used RPI datasets, RPI369, RPI488, RPI1807, RPI2241 and NPInter, RPITER obtained A U C of 0.821, 0.911, 0.990, 0.957 and 0.985, respectively. The proposed RPITER could be a complementary method for predicting RPI and constructing RPI network, which would help push forward the related biological research on ncRNAs and lncRNAs.

Sign in / Sign up

Export Citation Format

Share Document